Abidex

AbideX is zero-code monitoring for AI agents. Add one import before CrewAI, LangGraph, Pydantic AI and get spans with role, goal, and timing.

pip install abidex
pip install 'abidex[otlp]'       # OTLP → SigNoz / Jaeger / etc. (zsh: quote the extras)
pip install 'abidex[notebook]'   # optional: `abidex notebook` (Jupyter + pandas stack)

import abidex  # first
from crewai import Agent, Task, Crew
# ... run your crew

Traces go to console by default. For a persistent UI: abidex backend start then open http://localhost:8080 (SigNoz’s default; some older installs use 3301) and set OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317.

---

CLI

Traces vs logs. In OpenTelemetry these are different signals. Traces are made of spans: nested units of work (e.g. a CrewAI workflow, an agent step) with timing, parent/child links, and attributes such as gen_ai.agent.role. Use abidex trace / trace last and span exports like spans_*.ndjson. Logs are discrete log records (often from Python’s logging routed through the OTel logs SDK); they may include trace/span context, and Abidex can attach gen_ai.* fields from the active span when a line is emitted. Use abidex logs / logs last and files such as logs_*.ndjson. Both can be sent to the same backend (e.g. SigNoz) but the CLI treats span files and log files separately.

Command	Description
abidex run SCRIPT.py	Run a script with span buffer on, then print a trace table (-n / --last for row count)
abidex backend start	Clone SigNoz if needed (~/.abidex/signoz), docker compose up, open UI (8080 or 3301)
abidex backend stop	Stop SigNoz stack
abidex backend status	Check if SigNoz UI / OTLP ports are reachable
abidex status	Config, buffers, patched frameworks (-v for extra detail)
abidex trace · abidex spans	List span JSONL files (e.g. spans.ndjson, traces/, data/traces/)
abidex trace last [N]	Table of last N spans from buffer or latest export (--file, --filter, -v)
abidex trace export -f jsonl -o FILE	Export spans from buffer to JSONL (-f pretty for JSON stdout)
abidex logs	List log JSONL files (logs/, data/logs/)
abidex logs last [N]	Table of last N OTel log records (buffer or latest file; --file path.ndjson)
abidex logs export [-o FILE]	Export buffered logs to JSONL (default logs/logs.ndjson)
abidex notebook	Create notebooks/abidex_logs.ipynb from latest logs, optional date filter, launch Jupyter
abidex init	Print .env template and Docker one-liners
abidex summary	Span stats (count, duration, tokens; -v for more)

abidex backend start
abidex run main.py
abidex trace last 10
abidex logs last
abidex notebook

---

Docs

• Frameworks — CrewAI, LangGraph, Pydantic AI, LlamaIndex, n8n
• CrewAI integration — env vars, import order, troubleshooting
• Configuration — env variables
• Troubleshooting — common issues

---

Issues: GitHub · License: Apache 2.0